/* File:   ex4.18_pth_mat_vect_private.c  
 *
 * Purpose:  
 *     Computes a parallel matrix-vector product.  Matrix
 *     is distributed by block rows.  Vectors are distributed by 
 *     blocks.  This version uses private storage on each thread
 *     for its components of y.
 *
 * Input:
 *     m, n: order of matrix
 *     A, x: the matrix and the vector to be multiplied
 *
 * Output:
 *     y: the product vector
 *
 * Compile:  gcc -g -Wall -o ex4.18_pth_mat_vect_private 
 *                 ex4.18_pth_mat_vect_private.c -lpthread
 * Usage:
 *     ex4.18_pth_mat_vect_private <thread_count>
 *
 * Notes:  
 *     1.  Local storage for A, x, y is dynamically allocated.
 *     2.  Number of threads (thread_count) should evenly divide both 
 *         m and n.  The program doesn't check for this.
 *     3.  We use a 1-dimensional array for A and compute subscripts
 *         using the formula A[i][j] = A[i*n + j]
 *     4.  Distribution of A, x, and y is logical:  all three are 
 *         globally shared.
 *
 * IPP:    Exercise 4.18
 *         ff.)
 */

#include <stdio.h>
#include <stdlib.h>
#include <pthread.h>

/* Global variables */
int     thread_count;
int     m, n;
double* A;
double* x;
double* y;

/* Serial functions */
void Usage(char* prog_name);
void Read_matrix(char* prompt, double A[], int m, int n);
void Read_vector(char* prompt, double x[], int n);
void Gen_matrix(double A[], int m, int n);
void Gen_vector(double x[], int n);
void Print_matrix(char* title, double A[], int m, int n);
void Print_vector(char* title, double y[], double m);

/* Parallel function */
void *Pth_mat_vect(void* rank);

/*------------------------------------------------------------------*/
int main(int argc, char* argv[]) {
   long       thread;
   pthread_t* thread_handles;

   if (argc != 2) Usage(argv[0]);
   thread_count = atoi(argv[1]);
   thread_handles = malloc(thread_count*sizeof(pthread_t));

   printf("Enter m and n\n");
   scanf("%d%d", &m, &n);

   A = malloc(m*n*sizeof(double));
   x = malloc(n*sizeof(double));
   y = malloc(m*sizeof(double));
   
   Gen_matrix(A, m, n);
#  ifdef DEBUG
// Read_matrix("Enter the matrix", A, m, n);
   Print_matrix("We read", A, m, n);
#  endif

   Gen_vector(x, n);
#  ifdef DEBUG
// Read_vector("Enter the vector", x, n);
   Print_vector("We read", x, n);
#  endif

   for (thread = 0; thread < thread_count; thread++)
      pthread_create(&thread_handles[thread], NULL,
         Pth_mat_vect, (void*) thread);

   for (thread = 0; thread < thread_count; thread++)
      pthread_join(thread_handles[thread], NULL);

#  ifdef DEBUG
   Print_vector("The product is", y, m);
#  endif

   free(A);
   free(x);
   free(y);

   return 0;
}  /* main */


/*------------------------------------------------------------------
 * Function:  Usage
 * Purpose:   print a message showing what the command line should
 *            be, and terminate
 * In arg :   prog_name
 */
void Usage (char* prog_name) {
   fprintf(stderr, "usage: %s <thread_count>\n", prog_name);
   exit(0);
}  /* Usage */

/*------------------------------------------------------------------
 * Function:    Read_matrix
 * Purpose:     Read in the matrix
 * In args:     prompt, m, n
 * Out arg:     A
 */
void Read_matrix(char* prompt, double A[], int m, int n) {
   int             i, j;

   printf("%s\n", prompt);
   for (i = 0; i < m; i++) 
      for (j = 0; j < n; j++)
         scanf("%lf", &A[i*n+j]);
}  /* Read_matrix */


/*------------------------------------------------------------------
 * Function:        Read_vector
 * Purpose:         Read in the vector x
 * In arg:          prompt, n
 * Out arg:         x
 */
void Read_vector(char* prompt, double x[], int n) {
   int   i;

   printf("%s\n", prompt);
   for (i = 0; i < n; i++) 
      scanf("%lf", &x[i]);
}  /* Read_vector */

/*------------------------------------------------------------------
 * Function: Gen_matrix
 * Purpose:  Use the random number generator random to generate
 *    the entries in A
 * In args:  m, n
 * Out arg:  A
 */
void Gen_matrix(double A[], int m, int n) {
   int i, j;
   for (i = 0; i < m; i++)
      for (j = 0; j < n; j++)
         A[i*n+j] = random()/((double) RAND_MAX);
}  /* Gen_matrix */

/*------------------------------------------------------------------
 * Function: Gen_vector
 * Purpose:  Use the random number generator random to generate
 *    the entries in x
 * In arg:   n
 * Out arg:  A
 */
void Gen_vector(double x[], int n) {
   int i;
   for (i = 0; i < n; i++)
      x[i] = random()/((double) RAND_MAX);
}  /* Gen_vector */


/*------------------------------------------------------------------
 * Function:       Pth_mat_vect
 * Purpose:        Multiply an mxn matrix by an nx1 column vector
 * In arg:         rank
 * Global in vars: A, x, m, n, thread_count
 * Global out var: y
 */
void *Pth_mat_vect(void* rank) {
   long my_rank = (long) rank;
   int i, j;
   int local_m = m/thread_count; 
   int my_first_row = my_rank*local_m;

   double* local_y = malloc(local_m*sizeof(double));
	
   for (i = 0; i < local_m; i++) {
      local_y[i] = 0.0;
      for (j = 0; j < n; j++)
          local_y[i] += A[(i+my_first_row)*n+j]*x[j];
   }

   memcpy(y+my_first_row, local_y, local_m*sizeof(double));
   free(local_y);
   return NULL;
}  /* Pth_mat_vect */


/*------------------------------------------------------------------
 * Function:    Print_matrix
 * Purpose:     Print the matrix
 * In args:     title, A, m, n
 */
void Print_matrix( char* title, double A[], int m, int n) {
   int   i, j;

   printf("%s\n", title);
   for (i = 0; i < m; i++) {
      for (j = 0; j < n; j++)
         printf("%4.1f ", A[i*n + j]);
      printf("\n");
   }
}  /* Print_matrix */


/*------------------------------------------------------------------
 * Function:    Print_vector
 * Purpose:     Print a vector
 * In args:     title, y, m
 */
void Print_vector(char* title, double y[], double m) {
   int   i;

   printf("%s\n", title);
   for (i = 0; i < m; i++)
      printf("%4.1f ", y[i]);
   printf("\n");
}  /* Print_vector */